What about the findings from the ganzfeld meta-analysis done by Ersby, Maaneli and Johann (I think) looking at selected and unselected subjects, which was on a more complete pool. Weren't they considerably attenuated?
When the HR of the selected file-drawer was factored into the overall HR, it brought down the overall Storm (2010) hit rate of selected subjects from 40.1% to ~37%. The unselected file-drawer, however, confirmed the nearly universal ~26% hit rate for unselected participants, and had no effect upon the MA.
Subject selection alone does not guarantee a success rate of 40%; on the contrary, if participants are selected for only one or two traits, their HR is likely to be commensurate with the PRL database, at around 32%. The Storm et al (2010) selected HR of 40.1% reflects the improvement in participant selection from prior databases to more recent ones; stronger correlates have been used, or more of them.